Recently, Holden et al. [20] proposed a SNP-based pathway analysis, which used all available SNPs to represent a gene. This approach is computationally intensive and might not be practical for genome-wide studies with millions of SNPs. O'Dushlaine et al. [21] developed a SNP ratio test (SRT) method which computed the ratio of the number of significant (P < 0.05) to the number of non-significant (P >= 0.05) SNPs for each pathway and used permutations to identify the significant pathways. The SNP ratio test (SRT) method computes the ratio of the number of significant (P < 0.05) to the number of non-significant (P >= 0.05) SNPs for each pathway, and uses permutations to quantify its statistical significance. If there is only one gene that contains significant SNPs, the SRT method would reduce the pathway signal to a gene signal. By contrast, our method uses adaptive rank truncated product and permutations to determine the number of representative SNPs for each gene, and each gene includes at least one SNP. Therefore contributions from multiple genes are emphasized in the pathway analysis. Another disadvantage
